I include 4E in that. I have played adventures where skill challenges were run by the book--get X successes before Y failures, using the following skills--and it was the stupidest, boringest, immersion-wreckingest thing ever.
At the risk of a derailment, that's not "by the book". And in saying that, I'm not really meaning to take issue with you - your post is just the trigger for a more general point! - but to try to bring out an important aspect of non-combat action resolution.
What the books actually say to players is:
Your DM sets the stage for a skill challenge by describing the obstacle you face and giving you some idea of the options you have in the encounter. Then you describe your actions and make checks until you either successfully complete the challenge or fail. (PHB p 259)
It’s up to you to think of ways you can use your skills to meet the challenges you face. (PHB p 179)
And to GMs, they say:
Begin by describing the situation and defining the challenge. (DMG p 74)
More so than perhaps any other kind of encounter, a skill challenge is defined by its context in an adventure… Define the goal of the challenge and what obstacles the characters face to accomplish that goal… You describe the environment, listen to the players’ responses, let them make their skill checks, and narrate the results. (DMG pp 72, 73)
When a player’s turn comes up in a skill challenge, let that player’s character use any skill the player wants. As long as the player or you can come up with a way to let this secondary skill play a part in the challenge, go for it… In skill challenges, players will come up with uses for skills that you didn’t expect to play a role. Try not to say no… This encourages players to think about the challenge in more depth… However, it’s particularly important to make sure these checks are grounded in actions that make sense in the adventure and the situation… you should ask what exactly the character might be doing … Don’t say no too often, but don’t say yes if it doesn’t make sense in the context of the challenge. (DMG pp 73, 75)
I think this is all pretty straightforward. (The advice for setting DCs, on the other hand, remains wonky even through the Essentials books.) The GM describes a situation, the players describe how their PCs tackle it, checks are made based on that description, and the GM adjudicates the outcomes of those checks. The players then engage the new situation via their PCs, until the challenge either is overcome or overcomes, and the situation is thereby resolved (at X successes or 3 failures).
What is missing is a clear statement (i) of the rationale for the X succeses before Y failures structure, and (ii) of how to narrate the outcomes of checks so as to maintain the coherence of the fiction over the challenge. In my own case, I plugged those gaps by drawing on the better GMing advice found in other RPGs that have mechanics by which I assume skill challenges were inspired.
I also liked the concept of skill challenges but never managed to pull them off, it's an interesting concept that just don't work in practice
I don't agree with the end of your sentence. Skill challenges are a version of "extended contest" resolution systems seen in several systems, including HeroWars/Quest, Maelstrom Storytelling, and Burning Wheel's
Duel of Wits.
Those systems work in practice, and so do skill challenges provided similar techniques are used.
Here are some examples from my 4e game.
The rationale for an "X before Y" mechanic is much the same as the rationale for hit point depletion in combat: it gives a clear signal for when the conflict is resolved. (As opposed to relying on the mere intuition or fiat of one or more of the participants at the table.) I'm not saying that the skill challenge is the best or only way to do it - it is a "players roll all the dice variant", and that introduces certain oddities compared to the HeroQuest or Burning Wheel versions - but in its basic features it's a member of a well-established family of mechanics.
The tricky thing in any extended contest mechanic is to get the narration right - if you "close" too early, or leave matters "open" for too long, the narration and the mechanics can get out of whack. The standard solution to this problem is for the GM to metagame heavily in narrating the proceedings - ie in response to early successes, the GM introduces new, externally-sourced complications, and as the challenge is coming to an end, the GM has to be prepared to narrate in externally-sourced resolutions ("Ah, now I remember, you're the nephew of so-and-so who was my dear old friend . . ." would be a simple example of the latter).
The payoff is that (i) resoloving non-combat conflicts doesn't rely on GM fiat, and (ii) those conflicts occupy the same amount of time and dramatic "space" at the table as does combat, and hence contribute as much to the narrative. (I have seen it said that, because in a skill challenge the players can never win on their first skill check, no matter how brilliant their conception and execution, it is therefore a form of railroading. Key to running a skill challenge is realising that this mechanical impossibility is no different from the fact that, because of hit point rules, a PC
cannot kill a dragon on the first blow, no matter how brilliant the ambush and how florid the narration of the vicious strike to the beast's neck. The whole point of hit point based combat, and of skill challenges, is to create
extended contests. If you want quick combat, where one hit can kill, you use minions. If you want quick non-combat resolution, where one skill check can overcome a challenge, you use simple skill checks.)
Furthermore, the need for the GM to introduce complications as part of the process of resolution, and the need for the players to then respond to those complications, pushes the situation in unexpected directions. Interesting stuff happens, that no one wanted or had even thought of at the start, but that everyone is invested in by the time the conflict comes to an end. (This plays out differently from the compromise mechanic in a BW Duel of Wits, but performs something like the same functional role.)
Anyway, lessons for D&Dnext? If you want non-combat to be as significant a pillar as combat, you have to create resolution systems which (i) will make it have the same heft and depth at the table as combat, and (ii) will make players confident to invest PC-building resources into non-combat capabilities. Conversely, if non-combat action resolution is mostly single skill checks with huge amounts of GM fiat around the framing and the adjudication of the outcome, players have little incentive to invest PC-building resources (because no clear sense of what the payoff will be) and, when the chips are down, are likely to choose combat as a reliable and predictable resolution system, rather than relying on GM-fiated non-combat to get their PCs out of trouble.
There's other stuff that can be learned from skill challenges (and similar extended contest mechanics) too - like the importance of multi-dimensional stakes to encourage players to not always opt to engage via their highest numbers. A simple example from combat is the difference between melee and ranged - even a fighter with a fairly poxy ranged attack will still use that attack when his/her enemy can't be engaged in melee. Which is to say that combat can be framed in more dimensions than just "roll your attack skill and apply your damage". Similarly with non-combat - if you occasionally want the dwarf fighter to try to charm someone with CHA rather than intimidate them with STR, you need to frame situations where the stakes are multi-dimensional: for example, the player can see that his/her PC might be better off looking polite, even if the NPC ends up not being persuaded, than looking scary, even though that might be the easier route to persuasion.
This ties in to the GM metagaming the resolution, too: when the CHA check fails, for example, instead of the GM declaring that the dwarf tried to be charming but was in fact rude - which will just make the player resolve to always go for STR and intimidation - the GM narrates the dwarf being charming, but the NPC nevertheless declining to go along with the offer ("I would love to help, but sadly swore an oath to my now-dead father that precludes me from helping you in this endeavour" - that also opens up the door for another PC to try Religion, or even Forbidden Lore, to try to free the NPC from his/her oath!).
And it also helps if the stakes aren't always live or die - if the player knows his/her dwarf fighter will die unless the NPC helps, then no amount of desire to be seen as polite is going to override that! Using more subtle and nuanced stakes helps open up the space for a wider range of meaningful choices by players.
As far as its core stat check mechanics is concerned, D&Dnext is well-placed to support this sort of non-combat stuff. But it is going to need the mechanics to give it heft and pacing and freedom from mere fiat (some form of extended contest mechanics, even if not exactly skill-challenge style ones). And it is going to need better GMing advice than they were able to produce for 4e.
I'm personally not all that inspired by the playtest documents. They seem to canvass only single-dimensional stakes, of the live-or-die variety, and the treatment of interaction - both in the rule sections, and in the medusa encounter (the most interesting interaction in the module), is pretty pitiful in my view.
if you look at RPG's that remove the focus on combat and focus on different stuff, you get a fair number of D&D fans who will decry rather loudly that it's not even an RPG anymore.
I think this is right.
But the strange thing is, it seems to be much the same fans who decry balancing classes, and PCs more generally, around combat. It seems that they
want stuff other than combat to matter, but aren't interested in the well-known mechanical solutions that actually deliver that result!